ZeptoBars - RSS feed http://zeptobars.ru/en/ Microelectronics. Die-shots. Artificial intelligence. Lasers. en-us Tue, 10 Jun 2006 04:00:00 GMT Wed, 16 Apr 14 17:03:04 +0400 webmaster@zeptobars.ru 120 10 <![CDATA[Phillips PCF8574 - 8-bit I2C port expander : weekend die-shot]]> http://zeptobars.ru/en/read/NXP-PCF8574-i2c-port-expander
Sat, 12 Apr 14 22:50:43 +0400
<![CDATA[Fake audiophile opamps: OPA627 (AD744?!)]]> http://zeptobars.ru/en/read/OPA627-AD744-real-vs-fake-china-ebay Walking around ebay I noticed insanely cheap OPA627's. It's rather old, popular and high-quality opamps, often used in audiophile gear. Manufacturer (Texas Instruments / Burr Brown) sells them 16-80$ each (depending on package & options) while on ebay it's cost was 2.7$, shipping included.

Obviously, something fishy was going on. I ordered one, and for comparison - older one in metal can package, apparently desoldered from some equipment. Let's see if there is any difference.

Plastic one was dissolved in acid, metal can was easily cut:


Genuine TI/BB OPA627 chip first. We can see here at least 4 laser-trimmed resistors - now we see why it could cost that much. Laser trimmed resistors are needed here due to unavoidable manufacturing variation - parts inside opamps needs to be balanced perfectly.

"Chinese" 2.7$ chip. There is only 1 laser trimmed resistor, but we also notice markings AD (Analog Devices?) and B744. Is it really AD744? If we check datasheet на AD744 - we'll see that metal photo perfectly matches one in the datasheet .

What happened here?

Some manufacturer in China put an effort to find cheaper substitute for OPA627 - it appeared to be AD744. AD744 has similar speed (500ns to 0.01%), similar type (*FET), supports external offset compensation. AD744 also support external frequency compensation (for high-speed high-gain applications) but there was no corresponding pin on the OPA627 - so this feature is unused.

On the other hand AD744 has higher noise (3x) and higher offset voltage (0.5mV vs 0.1mV).

So they bought AD744 in the form of dies or wafers, packaged them and marked as OPA627. It does not seems they earned alot of money here - it's more of an economic sabotage. Good thing that they did not used something like LM358 - in that case it would have been much easier to notice the difference without looking inside...

Be careful when choosing suppliers - otherwise your design might get "cost-optimized" for you :-)

PS. Take a look at our previous story about fake FT232RL.]]>
Tue, 08 Apr 14 10:38:17 +0400
<![CDATA[SiTime SiT8008 - MEMS oscillator : weekend die-shot]]> http://zeptobars.ru/en/read/SiTime-SiT8008-MEMS-oscillator-rosnano SiTime SiT8008 is a programmable MEMS oscillator reaching quartz precision but with higher reliability and lower g-sensitivity. Also SiTime is one of companies who received investments from Rosnano - Russian high-tech investment fund.

Photo of MEMS die puzzled us for quite some time. Is it some sort of integrated SAW/STW resonator?

The trick is that to reach maximum Q-factor (up to ~186'000 according to patents) MEMS resonator must operate in vacuum. So they package resonator _inside_ the die in hydrogen atmosphere, then anneal it in vacuum so that hydrogen escapes through silicon. So we see here only a cap with contacts to "buried" MEMS resonator. We were unable to reach the resonator itself without x-ray camera or ion mill.

MEMS die size - 457x454 µm.

Thankfully relevant patents were specified right on the die : US6936491 US7514283 US7075160 US7750758 :)

Digital die contains LC PLL and digital logic for one-off frequency programming and temperature compensation.
Die size - 1409x1572 µm.

Poly level:

Standard cells ~250nm techology.
Sat, 29 Mar 14 02:54:16 +0400
<![CDATA[TI TL431 adjustable shunt regulator : weekend die-shot]]> http://zeptobars.ru/en/read/TI-TL431-adjustable-shunt-regulator-linear-supply Die size 1011x1013 µm.

Sat, 22 Mar 14 20:25:06 +0400
<![CDATA[TI LM393 - dual comparator : weekend die-shot]]> http://zeptobars.ru/en/read/TI-LM393-dual-comparator Die size 704x748 µm.

Tue, 18 Mar 14 13:34:46 +0400
<![CDATA[Ti TS5A3159 - 1Ω analog switch : weekend die-shot]]> http://zeptobars.ru/en/read/Ti-TS5A3159-analog-switch-SPDT Die size 1017x631 µm, 1µm technology.

Sun, 16 Mar 14 10:15:25 +0400
<![CDATA[FTDI FT232RL: real vs fake]]> http://zeptobars.ru/en/read/FTDI-FT232RL-real-vs-fake-supereal

Genuine FT232RL

After etching metal layers:

Let's take a closer look at different parts of the chip. Here are rows of auto-synthesized standard cells:



Fake FT232RL

This chip is completely different! We can notice right away that number of contact pads is much higher than needed. Chip has marking "SR1107 2011-12 SUPEREAL"

After etching metal layers:

Closer look at standard cells:

Different block of the chip has different look of standard cells. It is likely that some modules were licensed(?) as layout, not HDL:

First type of SRAM:

Second type of SRAM:

Finally - mask ROM programmed on poly level, so we can clearly see firmware data:

Comparison of manufacturing technology

ChipDie sizeTechnologySRAM cell area
FTDI FT232RL3288x3209 µm600-800 nm123 µm2
Fake FT232RL3489x3480 µm500 nm68 µm2 and 132 µm2

While technology node is comparable, it seems that original FT232RL used less metals, hence much lower logic cell density. Fake chip is slightly larger despite slightly more advanced technology.


It seems that in this case Chinese designers implemented protocol-compatible "fake" chip, using mask-programmable microcontroller. This way they only needed to redo 1 mask - this is much cheaper than full mask set, and explains a lot of redundant pads on the die. Fake chip was working kinda fine until FTDI released drivers update, which were able to detect fake chips via USB and send only 0's in this case. It was impossible to foresee any possible further driver checks without full schematic recovery and these hidden tricks saved FTDI profits.

What's the economic reason of making software fake of well-known chip instead of making new one under your own name? This way they don't need to buy USB VID, sign drivers in Microsoft, no expenses on advertisement. This fake chip will be used right away in numerous mass-manufactured products. New chip will require designing new products (or revisions) - so sales ramp up will happen only 2-3 years later. Die manufacturing cost is roughly the same for both dies (~10-15 cents) .

From now on one should pay more and more attention when working with small shady distributors. Their slightly lower price could cause numerous hours of debugging fun.]]>
Mon, 17 Feb 14 11:29:25 +0400
<![CDATA[ST 34C02 - 2048-bit EEPROM : weekend die-shot]]> http://zeptobars.ru/en/read/ST-34C02-dimm-spd-eeprom 2C interface, typically used as SPD chip in DIMM memory modules.
Die size 1542x1422 µm, 1.2µm half-pitch.

Fri, 17 Jan 14 20:10:21 +0400
<![CDATA[Microchip 24LCS52 : weekend die-shot]]> http://zeptobars.ru/en/read/Microchip-24LCS52-i2c-eeprom-charge-pump 2C interface.
Die size 1880x1880 µm, 2µm half-pitch.

Closer look at charge-pump:

Sun, 12 Jan 14 18:21:14 +0400
<![CDATA[Invensense MPU6050 6-axis MEMS IMU : weekend die-shot]]> http://zeptobars.ru/en/read/Invensense-MPU6050-6d-MEMS-IMU-gyroscope-accelerometer tried to take a photo of MPU6050. Now it's time for revenge!

Invensense MPU6050 is an integrated gyroscope and accelerometer with 16-bit readings. It contains 2 dies, soldered/welded face-to-face in multiple places (that's what was causing us troubles last time, required temperature for separation exceeded 600C).

On the overview photo you can see how not-flat they are. On a bigger die MEMS part is 28µm above surface, on smaller die - 100 µm above. Also, there is logic right under MEMS on the bigger die.

Size of big die is 2782x2718 µm, small die - 2778x2195 µm.

Small die, focus on top level. Width of smallest teethes is 1µm.
These teethes allow to sense their movement by change of capacitance between electrodes.

Small die, focus on bottom level:

Big die:

Below MEMS one can find conventional digital logic, ~250nm halfpitch

SRAM, cell area is 10.13 µm2:

Standard-cell-based logic:
Sun, 29 Dec 13 13:06:53 +0400